How We Test
The short version: we buy it, we use it for weeks, we measure what can be measured, and we write down what went wrong.
Most product recommendations on the internet are written by people who have never touched the product. We think that's the whole problem, so here is exactly what we do instead, in enough detail that you can hold us to it.
1. We buy everything, at retail
Every product we review is purchased by us, at the same price you would pay, from the same retailers you would use. We do not accept review units, loaners, long-term samples, or "gifted" products. We do not take sponsored placements, paid rankings, or brand partnerships of any kind.
This is expensive and it is the point. A publication that depends on manufacturers for hardware depends on manufacturers, and everybody involved knows it even when nobody says it. Paying retail is the only version of independence that survives contact with a bad review.
2. We test for weeks, not hours
Our minimum testing window is three weeks for wearables and small electronics, four to six weeks for appliances, and thirty nights for anything you sleep on. Most failures do not show up on day one. Motors get louder, apps drop off Wi-Fi after a firmware update, foam compresses, batteries fade, mop heads start to smell. You only find that by living with the thing.
3. We measure what can be measured
Where a claim can be tested, we test it rather than repeating it. That means particle counts logged at intervals for air purifiers, output measured against a resistive load bank for power stations rather than trusting a surge rating, GPS traces compared against a chest strap and a dedicated running watch, surface temperature by thermal probe for mattresses, decibel meters for anything with a motor, and controlled debris lines for vacuums.
Where a thing genuinely cannot be measured — whether a chair is comfortable, whether an interface is pleasant — we say so and describe it instead of inventing a number.
4. We read the one-star reviews
Before we publish, we read through the negative reviews on Amazon, the relevant subreddit, and manufacturer support forums, looking for the failure that repeats. A single angry review is noise. The same complaint appearing forty times across two years is a design decision, and it belongs in the article. Almost every "The catch" section on this site started as a pattern in somebody else's one-star review that we then went and reproduced ourselves.
5. We publish what we returned
Nine of the 41 products we bought this year went back. They don't get reviews — a takedown of an obscure product is content, not service — but the count is public and we update it, because a review site where everything turns out to be good is a review site that has stopped testing.
6. How scoring works
Scores are out of 10 and are assigned relative to the category and the price, not the universe. An 8.4 on a $1,099 vacuum and an 8.4 on a $170 vacuum both mean "very good for what it is and what it costs." We do not use half the scale: nothing we would not recommend gets published with a score at all, so if you see a number here, it cleared the bar. The interesting information is always in the drawbacks, not the digit.
7. What we get wrong
Sometimes a unit is faulty and we mistake it for a design flaw. Sometimes a firmware update fixes something we criticised. When that happens we update the article, date the update, and say what changed rather than quietly editing it. If you think we've got something wrong, tell us — we would rather be corrected than be consistent.
What we bought this week
One email, every Thursday. What arrived, what we kept, what went back, and the one thing we'd actually buy again. No sponsored slots, ever.
Free. Unsubscribe in one click. We never sell your address.